ABSTRACT
In recent years, commercial text-to-speech (TTS) products have become available. However, no current TTS can offer the best performance for all kinds of applications, so users have to select the most appropriate TTS for their particular need. Therefore, TTS system performance must be measured carefully. Because, as shown in Chapter 2, TTS is a highly integrated system of text analysis, phonetics, phonology and digital signal processing, total TTS performance can not be indicated by a single evaluation measure, but must be evaluated from various kinds of aspects. In this section, we will first look at examples of evaluation, and then show a proposed guideline for TTS evaluation.
